Aligning your data strategy to your business goals
AWS customers tell us that a lack of alignment between data projects and their company's goals usually lead to a misused, over-engineered data platform that delivers little value for the business. Low reusability of data assets, data inconsistency, poor data discovery, long wait times, and low data quality are typical complaints.
Common mistakes in building a data strategy include focusing too much on technical tools and trends, using edge tools, and missing the chance to accelerate business opportunities by providing business users with data that uses their own terminology, automating manual tasks for key metrics reporting, providing data quality visibility, and giving users autonomy for data exploration.
Your data strategy should focus on solving your business problems, such as performing better customer segmentation to increase conversion rates, improving customer satisfaction with personalization, reducing customer churn by anticipating retention actions, testing new products and new features faster with A/B tests to improve the customer experience, and any other strategies that can improve business or branding impact.
Companies frequently underestimate data governance. Most efforts in this area are in the analytics layer, and very few processes are automated. This generates an overhead to data engineering teams that have to understand the data and translate it to data consumers without understanding the business domain associated with the data. Data governance, when applied from data ingestion through data consumption, can empower data strategy. Processes that support rich data standardization, classification, and quality enable people to interact with data easily and gain access to it in an automated way.
Discovering your company's current stage
Moving a company from an entry stage of data usage maturity to a data-driven stage is difficult, because it requires capabilities, processes, and roles that can take time to implement. The following diagram presents different stages in data usage maturity.
Stage 1 (transactional). In stage 1, companies are focused on their core operations. They don't take advantage of the data around those operations, because they don't measure or use financial and operational performance indicators for their business. Today, we see very few companies at this stage. Most of these are startup companies in the early stages of their business.
Stage 2 (informed by data). In stage 2, companies use data to monitor their business health in terms of operational, financial, and departmental data that is analyzed inside each department in a siloed manner. Most enterprises that are at this stage have on-premises, proprietary systems, where sharing the data can be complex and expensive.
Moving stage 2 companies to AWS usually involves enabling them to extract, catalog, and share data between business areas, and start using advanced interactive analysis.
Stage 3 (based on data). Stage 3 includes companies that have already optimized their data usage. These companies use their data in different ways, depending on the industry:
-
Service companies such as financial services, healthcare services, ecommerce services and consumer packaged goods services know their customers' behaviors. They use data to create timely recommendations and offers based on these behaviors.
-
Manufacturing companies often use advanced forecasting analysis to optimize their production and supply operations.
-
Agricultural and manufacturing companies use data to optimize their logistics operations, improve process efficiency, and implement precision agriculture.
However, although the companies in stage 3 use data extensively, they require manual data analysis to take these actions.
Most companies today are in stage 3, although some of them use more advanced techniques such as machine learning (ML) models, and some are starting to experiment with advanced analytics.
Stage 4 (driven by data). Companies in stage 4 are already making decisions, often automatically, based on their data. However, this can be challenging. It requires confidence in the data and mechanisms in place for applications to use and react to the data. Stage 4 also requires data to be available for timely decision-making.
Automating two-way door decisions
Reversible (two-way door) decisions are great candidates for data-driven actions. For example, a company might decide to quarantine a product (stop selling it) after it receives negative reviews that represent a statistically high probability of product returns or customer complaints. The quarantine is reversible after the issue has been addressed, and the product can be placed back on sale.
Fraud detection is another example of a two-way, data-driven action. Companies might introduce mechanisms to avoid loss for their customers and platform, even if they encounter some false positives that have to be addressed. They can introduce improvements by measuring the results of current mechanisms and evaluating their effectiveness. After false positives are mitigated or validated by customers, transactions can be confirmed or retried by using two-factor authentication or similar process.
However, some actions aren't easily reversible and require further discussion and approval by a board of executives. These are called one-way door decisions. For example, actions that involve the construction of facilities or significant money investments are usually hard to reverse. These aren't good candidates for automatic data-driven actions.
A data-driven action should be evaluated for the visibility of its impact through constant measurement. These measurements help you decide to roll back a feature or to test and engage a team for deeper analysis of distinct behavior.